Head Pose Estimation Using Mediapipe and OpenCV

This project implements real-time head pose estimation using Mediapipe's Face Mesh and OpenCV. The program detects facial landmarks, estimates the orientation of the head in 3D space, and visualizes the head pose using axes overlaid on a live webcam feed.

---

Features

• Real-time face detection and landmark tracking using Mediapipe Face Mesh.
• Head pose estimation via 3D-2D point correspondence and cv2.solvePnP.
• Visualizes head orientation with overlaid X, Y, and Z axes.

---

Prerequisites

1. Python 3.7+
2. Libraries:
   • OpenCV
   • Mediapipe
   • Numpy

Install dependencies using:

pip install opencv-python mediapipe numpy

---

Usage

1. Clone the repository or save the code to a file.

2. Run the script:

   python headpose.py

3. Grant webcam access when prompted.

4. Observe the live head pose estimation with axes drawn on your face.

5. Press q to exit the application.

---

How It Works

1. Face Detection:
   • Mediapipe's Face Mesh detects 468 facial landmarks in real-time.
2. 3D Model Points:
   • Predefined 3D model points represent key facial landmarks (e.g., nose tip, chin, eyes).
3. Head Pose Estimation:
   • OpenCV's cv2.solvePnP computes the head's rotation and translation vectors.
4. Visualization:
   • The rotation is visualized by projecting 3D axes onto the image plane and overlaying them on the face.

---

3D Model Points Mapping

The following Mediapipe Face Mesh landmarks are used for head pose estimation:

Landmark	Index
Nose tip	1
Chin	199
Left eye corner	33
Right eye corner	263
Left mouth corner	61
Right mouth corner	291

---

Example Output

When the script is running, you will see:

• Red Line (X-axis): Indicates left-right tilt.
• Green Line (Y-axis): Indicates up-down tilt.
• Blue Line (Z-axis): Points outward, indicating the facing direction.

---

Customization

• Change Camera Resolution: Modify the camera_matrix to match your webcam's resolution.
• Use Different Landmarks: Adjust the model_points and get_image_points functions for additional landmarks.
• Add Features: Extend the project with gesture recognition or attention tracking.

---

Troubleshooting

• Webcam Not Detected: Ensure the webcam is properly connected and accessible.
• Low Accuracy: Check the camera_matrix and model_points values for accuracy.
• Dependencies Not Found: Ensure all required libraries are installed.

---

License

This project is licensed under the MIT License. See the LICENSE file for more information.

---

Enjoy experimenting with head pose estimation! 🎉